Yves here. Mirabile dictu! A VoxEU article discusses, admittedly in suitably dense economese, how economists create and enforce biases against taxation by using terminology that presupposes that it’s bad. And as Lambert noted after he saw it went up here: “That post got Tyler Cowen really ticked off, so it must be good.”
By Charles F Manski, Professor in Economics, Northwestern University. Cross posted from VoxEU
Economists usually think of taxation as inefficient. This column argues that the anti-tax rhetoric evident in much lay discussion of public policy draws considerable support from the prevalent negative language of professional economic discourse. Optimal income taxation doesn’t have to employ the pejorative concepts of inefficiency, deadweight loss and distortion; and this column argues that it is high time for economists to discard them and make analysis of taxation and public spending distortion-free.
The anti-tax rhetoric evident in much lay discussion of public policy draws considerable support from the prevalent negative language of professional economic discourse. Economists regularly write about the ‘inefficiency’, ‘deadweight loss’, and ‘distortion’ of income taxation.
• An influential article of Feldstein (1999) begins “The traditional method of analysing the distorting effects of the income tax greatly underestimates its total deadweight loss as well as the incremental deadweight loss of an increase in income tax rates”.
• In a recent review article in the Journal of Economic Literature, Saez, Slemrod, and Giertz (2012) state: “under some assumptions all responses to taxation are symptomatic of deadweight loss. Taxes trigger a host of behavioural responses intended to minimise the burden on the individual. In the absence of externalities or other market failure, and putting aside income effects, all such responses are sources of inefficiency”.
Students of economics learn that the formal usage of the concepts ‘inefficiency’, ‘deadweight loss’, and ‘distortion’ in normative public finance refer to a theoretical setting where a private economy is in competitive equilibrium and a government can use lump-sum taxes to modify the endowments of individuals.
In this setting, classical theorems of welfare economics show that any Pareto efficient social outcome can be achieved by having the government use lump-sum taxes to redistribute endowments and otherwise not intervene in the economy. Income taxes and other commonly used taxes logically cannot yield better social outcomes than optimal lump sum taxes but they may do worse. Deadweight loss measures the degree to which they do worse.
Careful instructors caution students that the theoretical setting envisioned in classical welfare economics is far removed from reality for a host of reasons, from the impracticality of lump-sum taxes to the presence of many forms of market failure.
Nevertheless, prominent applied public economists continue to take the theory quite seriously. For example, Feldstein (1999) concludes his article on the deadweight loss of the income by writing that “The analysis implies that a marginal increase in tax revenue achieved by a proportional rise in all personal income-tax rates involves a deadweight loss of two dollars per incremental dollar of revenue. This has important implications for the cost of financing incremental government spending”.
The Feldstein article and similar research on deadweight loss appear predestined to make income taxation look bad. The research aims to measure the social cost of the income tax relative to the utterly implausible alternative of a lump-sum tax. It focuses attention entirely on the social cost of financing government spending, with no regard to the potential social benefits. If applied economists are to contribute fair and balanced analysis of public policy, I think it essential that we jointly evaluate taxation and public spending within a framework that restricts attention to feasible tax instruments and that makes a reasonable effort to approximate the structure of the actual economy.
A constructive approach is to specify a social-welfare function and determine the welfare achieved by alternative feasible taxation and spending policies. An appropriate social-welfare function recognises both the costs of taxation and the benefits of tax-financed spending, making it unnecessary to invoke one-sided concepts such as inefficiency, deadweight loss, and distortion.
Indeed, Mirrlees (1971) made no mention of any of these concepts in his seminal study of optimal income taxation. In Mirrlees’ work and in the body of research stemming from it, the specified social-welfare function is the only normative concept required for evaluation of public policy. The welfare function and the specified set of feasible taxation and spending policies express the assumptions made about the structure of the economy.
Income Taxation and Infrastructure Spending
The gross inadequacy of deadweight loss and related concepts for policy analysis is particularly striking when considering tax-financed public spending for infrastructure that aims to enhance private productivity. Even as strong a proponent of private enterprise as Milton Friedman recognised the need for government to provide infrastructure for voluntary exchange, writing “In… a free private enterprise exchange economy, government’s primary role is to preserve the rules of the game by enforcing contracts, preventing coercion, and keeping markets free” (1955).
Friedman’s statement focuses on the need for government to provide laws, regulations, and a justice system to enforce them. One might reasonably add many further governmental functions, including formation and execution of monetary policy, provision and oversight of transportation and communications, protection of the environment, and support of research. Each of these and other functions may be performed with varying intensity, at correspondingly varying cost. Taxation is the main mechanism that governments use to finance infrastructure.
In recent work, I have studied optimal income taxation to finance public spending on infrastructure (Manski 2013). My analysis follows Mirrlees (1971) in some respects and differs in others. Following Mirrlees, I suppose that each member of a population allocates time to paid work and to the various non-paid activities that economists have traditionally called ‘leisure’. Persons have heterogeneous wages. An income-tax schedule is applied to gross income, yielding net income. Persons allocate time to maximise utility, which increases with net income and leisure. Social welfare is utilitarian. The chosen policy must balance the public budget, equating tax revenues and government spending.
I depart from the Mirrlees setup in several ways.
• First, I consider the use of tax revenue to finance public spending on infrastructure.
• Second, I suppose that persons may have heterogeneous preferences for income, leisure, and public spending.
• Third, the social planner may have partial knowledge of population preferences and of the productivity of infrastructure spending.
In contrast, the body of research on optimal income taxation stemming from Mirrlees (1971) has commonly studied the use of taxation to redistribute income while holding government spending fixed and has assumed that all persons have the same, known, preferences.
For the present discussion, the essential feature of my research is the transparent way that it characterises how public spending on infrastructure may enhance private productivity. I suppose that wages are person-specific positive constants multiplied by an aggregate production function expressing the wage-enhancing effect of infrastructure spending. This contrasts sharply with the common research practice of considering wages to be fixed.
Analysis of optimal policy is particularly simple in an illustrative setting that yields easily interpretable closed-form findings. In this setting, the planner only considers tax schedules that make the tax proportional to gross individual income. Persons have Cobb-Douglas income-leisure preferences and no non-labour income. These assumptions imply that time-allocation choices are invariant to policy.
The assumptions imply that optimal infrastructure spending is determined by the shape of the aggregate production function expressing the private productivity of public spending. Increasing the tax rate is socially beneficial if and only if the additional infrastructure spending enabled by the tax rise yields a more than commensurate increase in wages. The optimal public-spending level maximises aggregate net (after tax) income.
Performing this research, I found no need to use the concepts of inefficiency, deadweight loss, and distortion. Over 40 years ago, Mirrlees had it right when he found it unnecessary to use these pejorative ideas in his study of optimal income taxation. I think it overdue for the profession to discard them and make analysis of taxation and public spending distortion-free.